Samsung shifts half of HBM production to HBM4, suspends production of 8-layer HM3E.

```

Samsung Electronics is significantly shifting the strategic focus of its high bandwidth memory (HBM) production capacity forward, aiming to use its next-generation products to reverse its previous lag in the HBM competition against SK Hynix.

According to Korean media reports, Samsung Electronics has recently allocated half of its monthly HBM DRAM wafer input—about 75,000 out of approximately 150,000 wafers—to sixth-generation HBM4, with the remaining capacity allocated to 12-layer HBM3E. The 8-layer HBM3E, which has seen relatively weak demand, has been temporarily discontinued; its production capacity has been entirely shifted to HBM4. Under the premise of ensuring existing HBM3E order deliveries, Samsung has in effect concentrated all incremental production capacity on the HBM4 track.

This restructuring of production capacity has a clear competitive logic. In the HBM3E generation, Samsung missed out on a large number of Nvidia orders due to delayed quality certification, while SK Hynix secured the supply pattern by capitalizing on its first-mover advantage. In February this year, Samsung became the first in the world to achieve mass production and shipment of HBM4, generating approximately $1 billion (about 1.54 trillion KRW) in revenue within four months, with expected annual revenue reaching $10 billion.

On the demand side, HBM4 is also rapidly expanding. In addition to the demand tied to Nvidia’s next-generation AI accelerator Rubin, major cloud computing companies such as Google, Amazon, and Microsoft are accelerating the development of proprietary AI chips, further boosting the potential demand for HBM4 and giving Samsung a broader market space for expansion in this new track.

Catch-Up Strategy: Using HBM4’s First-Mover Advantage to Make Up for HBM3E Losses

Samsung’s setback in the HBM3E cycle is the direct reason for this strategic shift. SK Hynix completed Nvidia’s quality certification ahead of time and secured a share of the supply, while Samsung, hampered by certification delays, failed to win enough orders. Rather than continue to follow step by step on HBM3E, Samsung is now putting its focus on the HBM4, where it already enjoys a first-mover advantage—this is the logic behind Samsung’s current strategy.

From the downstream application perspective, HBM3E is mainly paired with Nvidia's current-generation AI accelerator Blackwell, whereas HBM4 will be used in the next-generation accelerator Rubin. While maintaining supply to existing HBM3E customers, Samsung has actually concentrated all additional production capacity onto the HBM4 line.

Samsung’s initial commercial results with HBM4 validate the potential of this path. After achieving the world’s first mass production and shipment of HBM4 in February this year, Samsung generated about $1 billion in revenue within four months, with a projected annual revenue of $10 billion.

ASIC Market Expansion: Samsung's Integrated Capability as a Potential Catalyst

Another important driver behind Samsung’s intensified push toward HBM4 is the rapid growth of the custom chip (ASIC) market. Major cloud computing companies such as Google, Amazon, and Microsoft are vigorously developing their own AI accelerators to reduce dependence on Nvidia GPUs, a trend that is driving rapid growth in demand for related HBM products.

Industry analysis suggests that, starting with the HBM4 generation, customized base dies and advanced packaging solutions tailored to different client needs will become more critical competitive factors. With its integrated layout across memory, foundry, and advanced packaging capabilities, Samsung has a differentiated advantage under this new competitive dimension.

SK Hynix: Holding Steady on the HBM3E Front, Transitioning Smoothly to HBM4

Compared to Samsung’s aggressive approach, SK Hynix’s position in the HBM market is more secure, and it does not need to rush into mass production of HBM4. The company has already established a strong supply position in the HBM3E market and is also expected to become Nvidia's largest supplier in the HBM4 stage.

Nvidia CEO Jensen Huang publicly stated during his visit to Korea this month that SK Hynix is its largest HBM supplier. A semiconductor industry insider noted, “There’s no rush; with the arrival of the Rubin production node, SK Hynix’s HBM4 revenue will naturally grow.” This means that SK Hynix can maintain stable deliveries of existing HBM3E demand while gradually shifting its focus to HBM4 at its own pace, making its overall transition much smoother than Samsung's.

According to forecasts by market research firm TrendForce and investment bank Bernstein, Samsung Electronics' HBM market share is expected to rise from 27% in 2025 to 37% in 2026, while SK Hynix’s share will drop from 56% to 43%. In addition, some analyses predict that Samsung may surpass SK Hynix in HBM market share as early as 2027.

As HBM4 capacity ramps up and demand in the ASIC market continues to grow, competition between Samsung and SK Hynix in the HBM market will further intensify. The outcome of this generational transition will largely depend on how well HBM4 capacity expansion matches downstream AI chip demand.

Risk Warning and DisclaimerThe market involves risks and investments require caution. This article does not constitute personal investment advice and does not take into account the individual investment objectives, financial situation, or needs of specific users. Users should consider whether any opinions, views, or conclusions in this article fit their particular circumstances. Investment is at your own risk. ```